PETScTAOSolver

class dolfin.cpp.la.PETScTAOSolver(tao_type='default', ksp_type='default', pc_type='default')

Bases: dolfin.cpp.la.PETScObject

This class implements methods for solving nonlinear optimisation problems via PETSc TAO solver. It supports unconstrained as well as bound-constrained minimisation problem

Create TAO solver for a particular method

static default_parameters()

Default parameter values

init()

Overloaded versions

• init(optimisation_problem, x, lb, ub)

  Initialise the TAO solver for a bound-constrained minimisation problem, in case the user wants to access the TAO object directly

• init(optimisation_problem, x)

  Initialise the TAO solver for an unconstrained minimisation problem, in case the user wants to access the TAO object directly

static methods()

Return a list of available solver methods

parameters

solve()

Overloaded versions

• solve(optimisation_problem, x, lb, ub)

  Solve a nonlinear bound-constrained optimisation problem

  Arguments

  optimisation_problem (OptimisationProblem)

  The nonlinear optimisation problem.

  x (GenericVector)

  The solution vector (initial guess).

  lb (GenericVector)

  The lower bound.

  ub (GenericVector)

  The upper bound.

  Returns

  num_iterations (std::size_t)

  Number of iterations

• solve(optimisation_problem, x)

  Solve a nonlinear unconstrained minimisation problem

  Arguments

  optimisation_problem (OptimisationProblem)

  The nonlinear optimisation problem.

  x (GenericVector)

  The solution vector (initial guess).

  Returns

  num_iterations (std::size_t)

  Number of iterations

• solve(optimisation_problem, x, lb, ub)

  Solve a nonlinear bound-constrained minimisation problem

  Arguments

  optimisation_problem (OptimisationProblem)

  The nonlinear optimisation problem.

  x (PETScVector)

  The solution vector (initial guess).

  lb (PETScVector)

  The lower bound.

  ub (PETScVector)

  The upper bound.

  Returns

  num_iterations (std::size_t)

  Number of iterations

tao()

Return the TAO pointer

thisown

The membership flag